Imaging Approaches in Cancer Biology

Abstract

A majority of cancer research is focused on defining the cellular and molecular basis of cancer cells and the signals that control oncogenic transformation; as a consequence, we know very little about the dynamic behavior of cancer cells in vivo. To begin to view and understand the mechanisms and interactions that control cancer initiation, growth, and metastatic progression and how these processes are influenced by the microenvironment and the signals derived from it, it is essential to develop strategies that allow imaging of the cancer cells in the context of the living microenvironment. Here, we discuss emerging work designed to visualize how cancer cells function within the microenvironment to discover how these interactions act coordinately to enable aberrant growth and to understand how they could be targeted to design new approaches to intercept the disease.

Imaging is a critical tool in the diagnosis and treatment of cancer. However, high-resolution visualization of the cellular and biological processes in normal and diseased states remains a challenge. Conventional imaging approaches such as positron emission tomography (PET) and magnetic resonance imaging (MRI) can capture a static image that allows for the longitudinal and noninvasive imaging of disease within a living organism, albeit at a lower resolution and without detailed spatial information (Wang et al. 2015; Huang et al. 2021), and can fail to truly capture the biology in real time. Recent and ongoing advances in in vivo imaging techniques such as multiphoton and single-photon confocal microscopy have offered novel approaches to view the dynamic behavior of cancer cells in real time during tumor progression (Wang et al. 2015) as these techniques enable tissue imaging at a cellular and subcellular resolution (Entenberg et al. 2018; Petersen et al. 2023). Therefore, in vivo imaging can be a powerful tool for better understanding important biological processes such as cell division, microenvironmental interactions, migration, and metastasis.

In this article, we highlight in vivo imaging techniques that are commonly used in normal development and cancer research with a predominant focus on fluorescence-based in vivo imaging (intravital imaging) in mouse models. This is one of the most extensively used technologies in preclinical studies and can be applied to understand not only the tumor cells’ behavior but also their interactions with cells of the microenvironment. Rapid advances in technologies such as microscopy, organoid culture, fluorescence labeling, and high throughout image processing have broadened the utility of in vivo imaging. These advances include the use of larger numbers of channels, the ability to image at greater depths and for extended periods of time, and the visualization of a wide range of tissues at vastly improved resolution. This has enabled the field to address challenging questions and explore uncharted territory. We will discuss examples of in vivo imaging applications, the discoveries they have fueled, and how these have fundamentally shifted our spatiotemporal understanding of cellular behavior and their responses to their surroundings in vivo in normal physiology and in disease.

OVERVIEW OF IN VIVO IMAGING APPROACHES

The use of mouse models has transformed our understanding of cancer initiation and progression as well as the definition of new molecular dependencies and their targeting in the context of drug development (Cheon and Orsulic 2011). However, the in vivo tracking of primary tumor growth and metastatic spread in animal models can often be challenging. Noninvasive in vivo imaging approaches provide an opportunity to monitor the growth and spread of tumors over time in living animals, generating valuable information about tumor burden, and the kinetics of tumor growth (Puaux et al. 2011). Most imaging techniques currently used in rodents are smaller versions of clinical imaging instruments used in the diagnosis and detection of human diseases. Table 1 lists some of the imaging modalities and summarizes their advantages and disadvantages.

Table 1.

Commonly used in vivo imaging techniques with their relative sensitivity, advantages, and disadvantages

Imaging scale	Technique	Brief overview	Advantages	Disadvantages
Whole body	Magnetic resonance imaging (MRI)	Uses powerful magnets to produce a magnetic field that forces protons in the body to line up A radiofrequency is pulsed through the body and this spins the protons in accordance with the properties of the water in that tissue MRI sensors detect the energy released as the protons realign with the magnetic field	Minimally invasive Good for longitudinal studies Distinguishes between soft tissues Has high spatial resolution Acquires physiological information No radiation	Small animal MRI is very expensive
Organs	X-ray (plain film and computed tomography [CT])	Uses electromagnetic radiation Images are compiled by a detector used to detect the X-rays that pass through a subject based on density and structural composition	Minimally invasive Good for longitudinal studies CT scans rotate around the subject increasing the depth of information gathered	Repeated exposure to radiation is dangerous Exogenous contrast agents to attenuate X-rays must be highly concentrated and have a high atomic number
	Ultrasound	Sound pulses are emitted from a transducer held against the subject and the reflected signal is used to construct an image Differences in the water/air content of tissues allow for the detection of different tissues Best for soft tissues	Made in vivo with only minor physiological impact Allows for accurate quantification of tumor volumes Minimally invasive Good for longitudinal studies	Difficulty in finding the optimal window to view organ or tumor of interest
Tissues	Positron emission tomography (PET)	Radiolabeled probes are injected into the subject The machine measures the positrons produced by the decay of the radiotracers that are emitted from the subject	Minimally invasive Good for longitudinal studies Measures biochemical properties of the subject Tracers exist to measure cell proliferation, bone remodeling, perfusion, oxygen metabolism, etc.	Requires radioactive materials
	Single-photon emission CT	γ-Emitting radioisotope is injected into the subject The γ rays are used to compile an image	Minimally invasive Good for longitudinal studies Accurate localization in 3D space Can use antibodies linked to a radioactive substance to label tumor cells	Requires radioactive materials
Cells	Bioluminescence imaging (BLI)	Genetically engineered bioluminescent reporter genes in cells/bacteria/viruses/viral vectors are transplanted into small animals Bioluminescent substrate is injected into the small animal Light signals are acquired Luciferase-/luciferin-based usually	High sensitivity, could even detect microscopic tumors None or low background noise Minimally invasive Good for longitudinal studies Can be combined with other imaging modalities for increased depth of knowledge Low cost, higher throughput Quantitative	Limited by use of single reporters and limited sensitivity in deep tissues Not transferable to large animals and humans
	Fluorescence-based “intravital” imaging (laser scanning confocal microscopy and multiphoton)	Uses fluorescent probes or fluorescent cells transplanted into mice Uses a laser to excite the fluorophores into emitting light	Both noninvasive and invasive imaging techniques Multicolor imaging capabilities for increased depth of knowledge Depending on the acquisition, can be a very high resolution Tracking cell behavior at a cellular/tissue level	Not easily translated to large animals and humans Limited depth of imaging

Bioluminescence imaging (BLI) is an imaging technique that is characterized by a high signal-to-noise ratio. It captures photons emitted by a luciferase (enzyme)-catalyzed oxidation of luciferin (exogenous substrate) (Paley and Prescher 2014). A wide range of bioluminescent enzymes, derived from a variety of organisms including insects, plants, and marine animals, have been adapted for use in mammalian cells (Lyons et al. 2013). Three of the most widely used luciferase–luciferin pairs for preclinical studies are derived from Photinus pyralis, Renilla reniformis, and Gaussia princeps. Each of these pairs can generate photons extracellularly, rendering them powerful resources for in vitro and in vivo studies (Hutchens and Luker 2007; Thorne et al. 2010). The substrate–enzyme pairs do not cross-react, permitting them to be used concurrently in a dual-labeling strategy to allow the monitoring of two different biological parameters sequentially and noninvasively in vivo. It is a highly sensitive technique that can detect signals from as few as 10 cells implanted in mice (Rabinovich et al. 2008), which makes it one of the most widely used in vivo imaging techniques in preclinical research. Despite its high sensitivity, the use of BLI is limited to qualitative assessment (i.e., detecting the presence or absence of a bioluminescent signal, rather than absolute quantitation) (Hutchens and Luker 2007; Badr 2014). In addition, widespread use of the most conventional luciferase–luciferin system is limited to a single reporter and has limited sensitivity in deeply located tissues (Mezzanotte et al. 2017). To overcome this limitation, different luciferin analogs have been developed that emit far-red to near-infrared bioluminescence and offer enhanced sensitivity and deeper tissue penetration (Jathoul et al. 2014; Kuchimaru et al. 2016).

MRI is a noninvasive procedure that generates detailed images of tissues within the body (Papanikolaou and Karampekios 2008). In brief, MRI renders images based on the spin-relaxation behavior of protons (¹H) in the aqueous environment when subjected to a strong magnetic field (Berger 2002; Lyons et al. 2013). Since the molecular and chemical composition of the environment influences how long it will take for protons to realign and the amount of energy they release during realignment, MRI signals can be used to distinguish healthy and diseased tissues. MRI differs from X-rays and computed tomography (CT) in that it also avoids the use of ionizing radiation, which can damage tissues, making it a valuable technique for imaging soft tissues such as the brain, spinal cord, muscles, ligaments, and tendons (Shah and Rojas 2023). However, MRI suffers from several disadvantages that include prolonged imaging time, higher cost, and a requirement for high technical expertise (Ladd et al. 2018). In preclinical settings, MRI has been commonly used for early tumor detection (Fig. 1), monitoring metastasis, and response to treatment (Peldschus and Ittrich 2014).

Figure 1.

Magnetic resonance imaging (MRI) in a mouse model of pancreatic cancer. MRI imaging of the pancreas in normal (left), tumor-bearing WT-KP^f/fC (middle), and Msi1^−/−-KP^f/fC (right) mice. (Figure reprinted from Fox et al. 2016, with permission.)

PET is a highly sensitive, noninvasive imaging tool that uses radioactive substances (tracers) to visualize the metabolic or biochemical function of the tissues within a body (Gambhir 2002). In this imaging technique, the administered radioactive tracers, such as 18-Fluorodeoxyglucose, accumulate in areas of the body with higher metabolic or biochemical activity, which helps determine the precise location of the disease. PET scans are much more sensitive than other optical imaging techniques such as CT and MRI and can thus detect abnormal metabolism in tumors before it can be detected by other imaging techniques. Therefore, it is well suited for preclinical and clinical imaging of various cancers. PET images are typically combined with CT as PET–CT or MRI as PET–MRI scans (Ehman et al. 2017; Yang et al. 2019), and are routinely used in clinical settings with applications in oncology, cardiology, and neurology. PET scans have also been widely used in preclinical studies, particularly for mouse models of cancer (Seemann et al. 2006; Duncan et al. 2013; Momcilovic et al. 2018). It has added advantages over CT and MRI, as the signal from the tracer can be readily and accurately detected without any impact from overlying tissue (Kinahan et al. 2003; Vaquero and Kinahan 2015; Nerella et al. 2022). On the other hand, PET imaging can only render and image with a low resolution and high noise (Kinahan et al. 2003).

Ultrasound (US) imaging is another noninvasive imaging technique that applies high-frequency sound waves (frequencies >20,000 Hz) to produce an image of internal organs (Cootney 2001). It has been extensively used in many areas of medical diagnostics. Because US imaging does not use ionizing radiation, it can be used for serial imaging without damaging the tissue under investigation (Wang et al. 2015). This feature makes it ideal for preclinical cancer studies that involve mouse models, as it can be used for early detection and to monitor and accurately measure tumor growth in autochthonous as well as transplant-based tumor models (Graham et al. 2005; Wirtzfeld et al. 2005; Sastra and Olive 2013). However, it also comes with limitations such as low sensitivity and low-resolution images with a narrow field of view (Ayers et al. 2010).

CT scans are noninvasive and use ionizing radiation (X-rays) to create 3D images of tissues within the body (Ferrero et al. 2019). In clinical settings, CT scans are routinely used to monitor tumor growth and assess the spread of cancer. In comparison to PET scans and most other imaging techniques, CT scans offer a much faster scanning and image processing capability (Vanderheyden 2009). CT is useful for tumor phenotyping and acquiring high-resolution anatomic details that provide context for the source of the imaging signals (Wang and Fleischmann 2018). However, its application can be limited in preclinical applications as it fails to provide functional information about the tissue under consideration, and it can be damaging to tumor cells due to the use of ionizing radiation.

OVERVIEW OF FLUORESCENCE-BASED IN VIVO (INTRAVITAL) IMAGING

The underlying principle of conventional fluorescence imaging centers on the excitation of fluorescent materials (i.e., dye [fluorescein, rhodamine, BODIPY, etc.], probe [Quantum dots], or reporter transgenes [GFP, YFP, RFP, etc.]) by a photon of a specific wavelength (Lyons et al. 2013). As a result, light with a longer wavelength is released and then detected (Sanderson et al. 2014; Elliott 2020). In contrast, multiphoton microscopy relies on excitation with longer-wavelength photons to produce emission with shorter wavelengths (Dunn and Young 2006; Svoboda and Yasuda 2006). Fluorescent materials can be selectively excited and visualized within living systems without compromising their biology. Fluorescence imaging offers superior results and advantages over other optical imaging modalities for understanding the biological process at a higher resolution. More importantly, the availability of a wide range of genetically encoded fluorescent proteins allows for the tagging of any protein of interest within a cell, permitting continuous tracking and measurement of spatial distributions and dynamics (Pittet and Weissleder 2011). These fluorescent proteins also allow for the labeling of specific and rare subpopulations of cells (Feng et al. 2000; Svoboda and Yasuda 2006).

Nonfluorescence-based imaging technologies such as PET and MRI provide valuable information and can be used to detect tumors and monitor response to therapy (Wang et al. 2015). However, these technologies typically cannot achieve subcellular to cellular resolution, and the data collected reflects static imaging. Similarly, imaging of fixed tissue sections or cells grown in vitro is informative but does not fully capture the complexity of the 3D structure of the tissues and the dynamic behavior of the cells in a physiological context in vivo. Recent advances in optical imaging techniques such as confocal imaging or multiphoton microscopy make it possible to visualize spatiotemporal dynamics of cells within a tissue in a living organism. In comparison to other techniques, such high-resolution optical techniques allow precise and repeated tracking of the same cells for a long duration of time, providing a better understanding of cellular dynamics during development and disease progression (Fumagalli et al. 2020a,b).

With the recent advances in multiphoton microscopy, deeply residing tumors labeled with fluorescent probes can be routinely imaged at a cellular or subcellular resolution (Entenberg et al. 2015; Huang et al. 2021; Scheele et al. 2022). Unlike conventional confocal microscopes, which use a pinhole aperture to maintain the plane of focus (Shimozawa et al. 2013), multiphoton microscopes use photons in a single plane of focus for nonlinear excitation (Larson 2011). In addition, as fluorophores are excited when they absorb multiple photons of longer wavelengths at the same time, multiphoton microscopy offers higher resolution and improves the signal-to-background ratio compared to conventional confocal microscopy, which uses photons with higher energy and shorter wavelengths. These critical differences make multiphoton imaging more efficient than conventional imaging techniques. This approach also allows for deeper penetration into tissues by reducing scattering and phototoxicity (Zipfel et al. 2003; Sidani et al. 2006). Thus, with the use of such high-resolution multiphoton microscopes, in vivo imaging has dramatically increased our understanding of how tumors grow, disseminate, and metastasize (Condeelis and Segall 2003; Condeelis and Weissleder 2010). However, regardless of the underlying principles, both single-photon and multiphoton microscopy have revolutionized our ability to visualize and track cellular behavior in vivo by allowing concurrent visualization of multiple cell types (both techniques are compatible with a wide range of currently available fluorescent reporters) and distinguishing transient and stable interactions between cells and their microenvironment (Koechlein et al. 2016). With the development of such techniques, in vivo imaging has enabled the in-depth study of tumor–stroma interactions along with their migratory behavior at a single-cell level, and has deepened our understanding of how the tumor microenvironment influences these processes (Rodriguez-Tirado et al. 2022; Duran et al. 2023; Genna et al. 2023).

COMMON APPROACHES TO ACCESS ORGANS/TISSUES TO PERFORM IN VIVO IMAGING

In vivo imaging is a technically challenging procedure. Despite the challenges, the adoption of in vivo imaging has exploded in recent years, mainly because of the significant technical advances in microscopy techniques, associated image processing software, and the development of a wide variety of labeling fluorophores (Entenberg et al. 2023). However, one of the most underappreciated factors in the widespread adoption of in vivo imaging comes from the technical innovation in surgical procedures that allows unhindered access to deeply residing tissues in the body under physiological conditions for long-term imaging while minimizing motion artifacts (Ritsma et al. 2013; Entenberg et al. 2023).

Easy and repeated access to tissues for imaging without repeated surgical intervention is important for longitudinal studies. Imaging windows can be used to provide repeated optical access to different organs for in vivo imaging (Alieva et al. 2014). The “cranial imaging window” provides a platform for live imaging in the brain and bone marrow. The “dorsal skin-fold chamber” (Lehr et al. 1993a,b; Jain et al. 2002) has been used for visualization of cancer cells and blood vessels in flank transplanted tumors. The “spinal chamber” permits visualization of spinal cord injuries and spinal tumors (Farrar et al. 2012). The “skin-flap technique” (Yang et al. 2002; Egeblad et al. 2008) is frequently used for accessible tissues such as skin, mammary glands, or subcutaneously implanted tumors. This technique, which is quick and easy to perform, involves making a small incision to expose the tissue or tumor to be imaged. The major limitation of these types of imaging windows is that animals cannot typically be imaged longitudinally in multiple sessions because tumors can dry out over time, rendering the mice susceptible to infection.

Permanently implanted imaging windows are a more recently developed technique. This approach is being more widely adopted because of its long-term advantages over conventional techniques, as it allows for imaging of the same tumor or cells for extended periods of time, days to weeks, in multiple imaging sessions. In this technique, the imaging window, which is usually made of titanium rings with a glass coverslip (Alieva et al. 2014), is permanently affixed (glued) to the imaging tissue and is sutured into the skin over the tissues of interest. Using this approach in the abdominal wall thus offers a novel way to visualize and track the spatial and temporal behavior of tumor cells in deeply located internal organs such as the small intestine, pancreas, liver, and kidneys (Ritsma et al. 2012, 2013). Similarly, an implantable optical window in the thoracic cavity enables lungs to be imaged over a prolonged period ranging from days to weeks (Entenberg et al. 2018; Borriello et al. 2020, 2021). Such imaging windows have provided novel insights into early tumor growth (Shimokawa et al. 2017; Maruno et al. 2021), dynamic interactions of tumor cells within their tumor microenvironment (Borriello et al. 2020, 2021, 2022), and response to therapy (Hirata et al. 2015). Overall, the use of imaging windows has significantly improved our ability to analyze dynamic cellular processes within deeply situated organs as they are extremely flexible and are highly compatible with most modern-day microscopes (Entenberg et al. 2023).

APPLICATIONS OF IN VIVO IMAGING IN BIOMEDICAL RESEARCH

In vivo imaging techniques can be useful for studying longitudinal dynamics and cell behavior within intact tissues, including tumors in living mice. In vivo imaging has been widely used in fundamental research to study a range of cellular activities including migration, intravasation, and cell fate such as mitosis, apoptosis, and division (Condeelis and Weissleder 2010).

IMAGING NORMAL STEM AND PROGENITOR CELL BEHAVIOR

Our current appreciation for high-resolution live imaging techniques has been derived from seminal studies involving the application of powerful in vivo imaging to visualize the dynamic interactions and organization of murine leukocytes in their physiological environment within intact lymphoid tissues (Miller et al. 2002; Stoll et al. 2002). While the existence and ability of leukocytes to move, interact, and extravasate from blood vessels to tissues were well documented and recognized for many decades, our ability to apply confocal and multiphoton microscopy to visualize cells at a high resolution in vivo has enabled us to follow the trajectory of cell movement and gain a better understanding of how their interactions contribute to an effective immune response (Germain et al. 2012). Besides the fundamental understanding of immune cell behavior, in vivo imaging has also been used to explore the dynamic interactions between host immune cells and pathogens (Coombes and Robey 2010). Currently, in vivo imaging is being used extensively to understand fundamental behaviors and interactions of a wide range of cell types such as stem cells and neurons with their niche cells in normal and disease contexts.

Longitudinal in vivo imaging of hematopoietic cells in the bone marrow has been used to define the stem cell niche. In one study, fluorescently labeled stem cells were transplanted to probe the interactions between individual hematopoietic stem and progenitor cells (HSPCs) and the vasculature, osteoblasts, and endosteal surfaces during their homing and engraftment in recipient mice (Lo Celso et al. 2009). This study was the first to demonstrate that asynchronous division of the implanted HSPCs is an important first step during efficient engraftment in the bone marrow endosteum. Moreover, the study was important in helping us understand how hematopoietic and microenvironmental cells interact within the niche and demonstrated that where a cell localizes depends on its differentiation stage, with long-term HSCs residing in closer proximity to endosteum and osteoblasts than more mature progenitors.

The dynamic behavior of stem and progenitor cells in vivo has also been studied using upright confocal microscopy, which allows multichannel imaging with greater light recovery. This approach is highly effective when depths beyond a few hundred microns are not essential. Using this approach, Koechlein et al. (2016) mapped the behavior of HSPCs in the bone marrow microenvironment (Fig. 2A). In this study, investigators used transgenic mice expressing a fluorescent protein (dsRed) throughout the body to label the bone marrow microenvironment. This approach allowed encounters between dsRed⁺ niche cells and hematopoietic cells labeled with a different fluorescent tag (GFP) to be visualized and analyzed at a single-cell level. Tracking of transplanted GFP-labeled hematopoietic stem cells revealed preferential localization and more stable associations with the vascular niche compared to the endosteum (Fig. 2B). In contrast, more differentiated cells made fewer stable associations with the vascular domain and tended to move more frequently between the vasculature and endosteum. These findings suggest that the differentiation status of a cell regulates its spatiotemporal behavior and dictates the types of signals it receives within the microenvironment. Overall, these studies demonstrate that in vivo imaging can be a powerful tool for gaining new insight into the dynamic behavior of hematopoietic cells in their native microenvironment. This strategy has also proven very useful in tracking individual leukemia cells and understanding their behavior in the context of hematologic malignancies (Fig. 3A).

Figure 2.

In vivo imaging of the bone marrow to study interactions between hematopoietic cells and the niche. (A) Representative image of mouse calvarium for in vivo imaging. Red-dashed boxes indicate imaging area. Black-dashed box shows the central sinus flanked by parasagittal sinusoids. (B) GFP⁺ hematopoietic stem/progenitor cells were transplanted into recipient mice; high-throughput computational analysis was used to monitor the movement of individual cells (green) over time, and determine their distance from elements in the niche including vascular (outlined in red) and endosteal (outlined in gray) regions. (Panels are reprinted from Koechlein et al. 2016, under Creative Commons Attribution 4.0 International (CC BY 4.0) License.)

Figure 3.

Application of in vivo imaging to study the interaction of normal/tumor cells with their niche. (A) GFP⁺ bcCML forms strong interactions with the dsRed-actin bone marrow niche cells. (B,C) Live images of calvarial bone marrow of dsRed mice transplanted with (B) shControl or (C) shSdc1 bcCML; shRNA vector (green), VE-cadherin (blue), Actin dsRed (magenta), white arrowheads show cells within the vasculature, red arrowheads show cells outside of vasculature within the marrow, yellow arrowheads show cell clusters. (D–F) Live animal imaging of dsRed recipient mice transplanted with (D) normal GFP⁺ hematopoietic stem/progenitor cells, (E) CD98^+/+ AML, or (F) CD98^−/− AML; transplanted cells (green), VE-cadherin (blue), Actin dsRed (magenta), yellow arrowheads point to blood vessels, white arrow marks a leukemia cell interacting with a blood vessel. Scale bar, 25 μm. (Panels A–C reprinted from Spinler et al. 2020, under Creative Commons Attribution 4.0 International (CC BY 4.0) License. Panels E–F reprinted from Bajaj et al. 2016, with permission from Elsevier © 2016.)

In vivo imaging, specifically multiphoton microscopy, has also been a valuable tool in the field of neurobiology, and has contributed to our current understanding of neurobiological phenomena, such as neurogenesis, neural differentiation, and the dynamics and functional organization of individual synapses during adulthood. Pilz et al. (2018) investigated the process of hippocampal neurogenesis using long-term in vivo imaging through the cranial window. This study revealed that nonradial, glia-like neural stem cells undergo a limited number of asymmetric divisions before their disappearance in adult mice, and provided a deeper understanding of how neural stem and progenitor cells transition from a proliferative to a neurogenic state during hippocampal neurogenesis. This approach has also provided insight in the case of ventricular–subventricular zone (V-SVZ) neurogenesis, where neural stem and progenitor cells were shown to sustain V-SVZ neurogenesis by limited rounds of symmetric division and self-renewal throughout life (Obernier et al. 2018). In vivo imaging through a cranial window has also been used to understand the dynamics of the invading edge of a glioblastoma tumor, where cancer stem cells (CSCs) can move differentially between the bulk of the tumor and the normal brain parenchyma (Alieva et al. 2019). Understanding these properties and how they relate to the ability of CSCs to initiate a new tumor can improve resection techniques and the accuracy of prognosis.

Sequential longitudinal in vivo imaging has also been effectively used to track the clonal dynamics of stem cells during intestinal homeostasis at the single-cell level. By imaging and tracking the progeny of Lgr5⁺ stem cells in Lgr5-confetti (Lgr5^{EGFP-Ires-CreERT2}/R26R-Confetti) mice in vivo, Ritsma et al. (2014) demonstrated a remarkable heterogeneity in the self-renewing potential of adult intestinal stem cells. The authors demonstrated a significantly higher long-term self-renewal potential of Lgr5⁺ intestinal central stem cells than their adjacent cells in a normal physiological state. This study suggests that only a small fraction of Lgr5⁺ cells possess true stem potential.

IN VIVO IMAGING IN HEMATOLOGIC MALIGNANCIES

In vivo imaging has been a valuable resource for understanding the establishment and progression of leukemia in the bone marrow. This allows for prolonged monitoring of the same region with tremendous spatial and temporal resolution previously not achieved by conventional static imaging techniques. High-resolution in vivo imaging has allowed researchers to track the movement, interactions, and distribution of leukemia cells in bone marrow in real time, which has helped identify key molecular partners and surface receptors that play a role in the maintenance and survival of leukemia cells. In vivo imaging has also been used to study the interactions between leukemia cells and stroma in the bone marrow and determine how these interactions are disrupted by therapy. Ex vivo live imaging has also been used as an alternative to in vivo imaging to observe leukemia cells in the bone marrow niche to understand biological phenomena, such as identifying the niche cells protecting therapy-resistant acute lymphoblastic leukemia (ALL)-propagating cells in response to chemotherapy (Duan et al. 2014). However, this approach does not fully recapitulate the physiologic parameters of the bone marrow and limits the dynamic range of information that can be gathered.

A sequential in vivo imaging approach using an ALL mouse model showed that leukemic cell growth significantly alters the bone marrow niche for normal HSPCs. These malignant cells create abnormal microenvironments that disrupt the natural homing and functioning of normal HSPCs (Colmone et al. 2008). However, contrary to the widely accepted hypothesis that cancer cells must interact closely with their microenvironment to survive, in vivo imaging revealed that human T-cell acute lymphoblastic leukemia cells (T-ALL) implanted into mouse bone marrow are highly migratory and do not exhibit preferential association with a particular cellular component of the bone marrow microenvironment (Hawkins et al. 2016). Similarly, in vivo imaging has also been used to compare and characterize the migratory behavior of different subtypes of leukemia cells and define how this behavior changes after chemo- and targeted therapy. In a murine AML model, treatment-naive leukemia cells were shown to be highly migratory, whereas surviving leukemia cells become less migratory in response to chemotherapy. In contrast, the leukemia cells in Notch-1-driven T-ALL remained highly motile even after chemotherapy (Duarte et al. 2019). Further, in contrast to AML, the ability of T-ALL cells to interact with and explore their environment in vivo is highly dependent upon chemokine receptor signaling (CXCR4). Overall, these in vivo imaging studies have revealed fundamental differences in the behavior of distinct leukemia lineages in vivo and provided important mechanistic insight into pathways regulating their interactions with the microenvironment (Duarte et al. 2019).

Using in vivo imaging, therapy-resistant leukemia stem cells in a BCR-ABL/NUP98-HOXA9-driven model of blast crisis chronic myeloid leukemia (bcCML) were shown to interact closely with the bone marrow niche. These interactions were shown to be mediated by Syndecan-1 (Sdc1)-dependent integrin-β7 (Itgb7) signaling (Fig. 3B,C). Disruption of Sdc1/Itgb7 signaling significantly impacted the survival and migration of bcCML cells in vivo (Spinler et al. 2020). Like bcCML, leukemia cells in MLL-AF9/NRAS^G12Vdriven AML tumor models have been shown to interact with their bone marrow niche in a CD98 signaling-dependent manner for their growth and survival in vivo (Fig. 3D–F). The disruption of CD98 interacting with the bone marrow microenvironment significantly reduced the long-term stable interactions of leukemia cells with their niche, specifically endothelial cells in vivo. These findings imply that the growth and dissemination of leukemia critically rely on adhesive signals and dynamic interactions with the niche and exemplify how in vivo imaging can be used to reveal key insight into how the microenvironment supports the growth of aggressive, therapy-resistant leukemia stem cells.

IN VIVO IMAGING IN SOLID CANCERS

In addition to its widespread use in studying the dynamics of cells in various tissues during normal developmental processes, in vivo imaging has been used to visualize interactions between tumor cells and elements in the microenvironment in solid cancers. Before the introduction of fluorescent reporter proteins, the use of in vivo imaging was predominantly limited to studying the short-term behavior of tumor cells in the context of their interactions with the vasculature (Suzuki et al. 1996; Scherbarth and Orr 1997). However, the use of long-term imaging windows in combination with genetic labeling of cells with fluorescent reporters has allowed researchers to expand their investigations into many important biological processes in vivo. For example, in vivo imaging has been used to study migratory behavior at primary sites (Condeelis and Segall 2003), understand the mechanism of metastatic dissemination (Kedrin et al. 2008; Robinson et al. 2009), and identify the fate of tumor cells in distant metastatic sites (Albrengues et al. 2018; Dai et al. 2022).

The vast majority of cancer deaths are due to the metastatic spread of the disease (Mehlen and Puisieux 2006). Therefore, a detailed understanding of the processes and mechanisms controlling tumor cell migration and seeding during metastasis is key to therapeutic success in cancer. This is especially true considering the fact that tumor cells are able to escape the confines of their preneoplastic lesions and undergo intravasation well ahead of their detectable limit (Harper et al. 2016). In this regard, in vivo imaging has significantly informed our current understanding of tumor metastasis. In a seminal study (Wang et al. 2002), Condeelis and colleagues used multiphoton in vivo microscopy to directly compare the migratory behavior of metastatic and nonmetastatic breast cancer cells. This study showed that metastatic and nonmetastatic cancer cells move at a similar rate in vivo and in vitro. However, striking differences were observed with their movement in vivo, where only metastatic tumor cells had the ability for directional motility on collagen-containing extracellular matrix fibers converging toward blood vessels. This study was instrumental in broadening our current understanding of tumor cell behavior during their migration in vivo.

In vivo imaging has also been used successfully to demonstrate the direct involvement of niche cells such as tumor-associated macrophages during intravasation and metastasis. Using a two-fluorescent reporter system in a metastatic mammary tumor model, investigators elegantly showed that perivascular macrophages directly interact with tumor cells to create cell clusters near the vasculature to initiate their collective movement and entry into the blood vessels (Fig. 4). This study offered early insight into how tumor cells migrate and enter the circulation in association with tumor-associated macrophages (Wyckoff et al. 2007). Subsequent studies involving in vivo imaging of the tumor microenvironment in breast cancers have been influential in establishing a new paradigm that tumor cells at the primary site intravasate exclusively through tumor microenvironment of metastasis (TMEM) doorways (Wyckoff et al. 2007; Robinson et al. 2009; Borriello et al. 2020; Sharma et al. 2021), portals that include perivascular macrophages, tumor cells, and endothelial cells (Fig. 4).

Figure 4.

Application of in vivo imaging to understand tumor metastasis. Intravital imaging was used to capture the intravasation process of cancer cells at primary sites. Primary cancer cells (breast cancer in this example) with metastatic potential crawl along the collagen fibers in the extracellular matrix (blue) that mostly converge to blood vessels. A paracrine signaling loop between macrophages (which release the cancer cell chemoattractant epidermal growth factor [EGF]) and cancer cells (which release the macrophage chemoattractant colony-stimulating factor-1 [CSF1]) is created, which results in amplification of the signal, resulting in their mutual attraction. Macrophages in direct contact with endothelial cells cause transient vascular openings termed TME of metastasis (TMEM) doorways (triangle), resulting in single cancer cell intravasation through the doorway. (Figure adapted from Condeelis and Segall 2003, with additional data from Wyckoff et al. 2007 and Robinson et al. 2009.)

In addition to the role of macrophages, in vivo imaging has been influential in providing visual evidence for the role of other immune cells such as neutrophils in tumor metastasis and dormancy. Egeblad et al. showed how sustained inflammation can play an important role in the reawakening of dormant cancer cells at a distant metastatic site (Albrengues et al. 2018). In a metastatic mammary tumor model, investigators showed that intravenously injected tumor cells could readily metastasize to distant sites such as the lungs but remain quiescent and survive for a prolonged period without further expansion. However, local inflammation forces these quiescent cells out of their dormancy. In this study, the authors showed that the inflammation caused excess accumulation of neutrophils and the release of protease-rich neutrophil extracellular traps (NETs). Proteases within the NETs, including neutrophil elastase (NE) and matrix metalloproteinase-9 (MMP9), acted to sequentially cleave laminin, leading to integrin activation and proliferation of quiescent cancer cells (Fig. 5). This study was influential in our understanding of how dormant tumor cells reactivate out of quiescence and the role of immune cells in this process.

Figure 5.

Application of in vivo imaging to understand the process of tumor dormancy and metastasis. Cancer cells from a primary tumor (breast cancer in this example) can disseminate to distant sites such as the lung and remain dormant for a prolonged duration (left). Inflammatory conditions such as chronic smoking can cause infiltration of neutrophils to metastatic sites. These neutrophils release neutrophil extracellular traps (NETs) laden with proteases including neutrophil elastase (NE) and MMP-9 (right). Protease activity leads to cleavage of the extracellular matrix protein laminin and exposure of critical epitopes that trigger integrin signaling and reactivation of dormant cancer cells. (Figure adapted from Albrengues et al. 2018.)

In addition to the use of in vivo imaging to visualize the movement and behavior of tumor cells during intravasation at the primary tumor site, it is also being used to study the fate of disseminated tumor cells (DTCs) at distant metastatic sites. This application allows researchers to visualize the interaction of tumor cells in a new microenvironment to understand their early fate, the drivers of dormancy, and their reawakening from dormancy leading to metastasis. In a recently published study involving a metastatic mammary tumor model (Borriello et al. 2022), investigators used a high-resolution in vivo imaging technique to track the behavior and fate of DTCs in lungs for several weeks. The study showed that, spontaneously, DTCs extravasate faster, survive for a greater period of time, and stay dormant significantly longer in the lung compared to intravenously injected metastasized tumor cells (Borriello et al. 2022). This work demonstrated that the primary tumor microenvironment can have a significant impact on the fate of tumor cells, affecting their dissemination, survival, and dormancy at distant metastatic sites. Likewise, in vivo imaging has enhanced our understanding of the mechanisms of tumor dormancy and their reawakening in vivo. In a study using a metastatic triple-negative breast cancer model, Dai et al. (2022) showed that brain metastasized DTCs preferentially reside within a vascular niche by directly interacting with perivascular astrocytes to stay dormant for a prolonged period before their regrowth. They also showed that astrocytic laminin (laminin-211) dictates cancer cell dormancy, as these specialized astrocytes suppressed the growth of dormant cancer cells in the brain. Similarly, in a separate study involving a metastatic mammary tumor model, disseminated single dormant tumor cells were imaged in the lungs to understand the mechanism of tumor cell dormancy and the mechanism of their awakening (Albrengues et al. 2018). This effort led us to appreciate the critical role of inflammation and NETs (Fig. 5) in the awakening of dormant cancer cells (Albrengues et al. 2018).

Another important utilization of in vivo imaging in solid cancers has been to visualize and demonstrate CSC activity in the primary tumor and during metastasis. In a genetically engineered mouse model of pancreatic ductal adenocarcinoma (PDAC), pancreatic cancer stem/progenitor cells, marked by Musashi (Msi2) reporter activity in a Msi2-GFP knockin reporter mouse and a highly aggressive therapy-resistant state, were shown to be located in spatially restricted clusters with some spatial bias toward vasculature within tumors (Fig. 6A,B; Fox et al. 2016). In vivo imaging of the pancreas in an autochthonous tumor model was used to demonstrate that Dclk1⁺ tumor stem cells generate new progeny within PDAC through longitudinal imaging from precancerous lesions to PDAC (Maruno et al. 2021). In the context of breast cancer, in vivo imaging of a murine mammary tumor demonstrated the dynamic and plastic nature of CSCs, where the preexisting CSCs are continuously replaced by newer CSCs throughout tumorigenesis (Zomer et al. 2013).

Figure 6.

In vivo imaging of the pancreas. (A,B) Live images of Musashi (MSI) reporter cells in pancreatic tumors from (A) REM1-KP^f/fC (Msi1-reporter) and (B) REM2-KP^f/fC (Msi2-reporter) mice; VE-cadherin (magenta), Hoechst (blue), MSI reporter (green). (C,D) In vivo imaging of tumor vasculature marked by VE-cadherin (magenta) within the pancreas of KP^f/fC mice treated with (C) vehicle or (D) the RORγ inverse agonist, SR2211. (E,F) In vivo imaging of autochthonous tumor-bearing REM2-KP^f/fC mice treated with (E) vehicle or (F) SR2211; Msi2-reporter (green), VE-cadherin (magenta), Hoechst (blue). (Panels A and B are reprinted from Fox et al. 2016, with permission from the authors. Panels C–F are reprinted from Lytle et al. 2019, with permission from Elsevier © 2019.)

CSCs have long been considered the cells responsible for fueling primary tumor growth and seeding distant metastasis in colorectal cancers. In vivo imaging has helped investigators explore the metastatic cell of origin in colon cancer, shedding new light on the inherent plasticity of CSCs and their role in metastatic colonization. Using orthotopic transplantation of tumor cells from a genetically engineered mouse model, Van Rheenen and colleagues demonstrated that Lgr5-negative (Lgr5⁻) cancer cells make up the bulk of circulating tumor cells as well as those at metastatic sites and drive the growth of metastatic tumors in colon cancer (Fumagalli et al. 2020,b). Through live imaging studies, the investigators demonstrated the intrinsic capacity of Lgr5⁻ cells to convert to Lgr5⁺ cells, by tracking the conversion of the fluorophore from RFP (marking all cancer cells) to eGFP (marking Lgr5⁺ CSCs) in vivo. They also showed that Lgr5⁺ cells drive the outgrowth of the tumor, but not the initial establishment of metastases.

In vivo imaging has also been used to study human cancers. While studies of human disease in vivo can be challenging, one of the most useful systems for studying the dynamic behavior of human cancer cells in living animals involves human organoids. Organoids are 3D cellular structures (mini organs) grown in vitro to mimic organ structures. Because they preserve the molecular and cellular characteristics of primary human tumors including genomic and mutational signatures over many passages, they have been widely adopted as a model system for studying key biological activities and response to therapy ex vivo (Wensink et al. 2021). Importantly, organoids maintain their phenotype when transplanted into mice, and can be easily engineered to express fluorescent reporter molecules, thus providing an ideal system for live imaging of human disease (Fumagalli et al. 2017, 2020a,b). In a powerful example of this approach, Sato and colleagues performed in vivo lineage tracing on patient-derived colorectal cancer organoids and tracked the fate of fluorescently labeled LGR5⁺ cells after transplantation. Through this, they showed that LGR5 marks the CSC population, and that LGR5⁻ non–stem cells harbored some level of plasticity (Shimokawa et al. 2017). Further, using a similar in vivo imaging approach in a subsequent study, they showed that a subpopulation of these CSCs, enriched in p27, can remain dormant and persist after conventional chemotherapy (Ohta et al. 2022).

Collectively, high-resolution in vivo imaging has not only helped us visualize cellular dynamics in the steady state but has also allowed us to monitor the real-time impact of therapies on tumor cells as well as the tumor microenvironment (Fig. 6C–F; Lytle et al. 2019). This has allowed investigators to assess cellular responses to anticancer therapies, including evaluation of cell death, tumor–stroma interactions, and drug resistance, and has provided important insight into the mechanism of therapy resistance in vivo (Nakasone et al. 2012; Arlauckas et al. 2017; Rodell et al. 2018).

CURRENT LIMITATIONS OF IN VIVO IMAGING AND FUTURE PERSPECTIVES

Although significant advances have been made since the first reported application of in vivo imaging, several challenges still exist, including the lack of availability of a wide range of compatible fluorophores to permit multiple colors to be imaged simultaneously, narrow imaging window, shallow penetration depth, limitations in optical access to tumors based on tumor site, and speed of imaging and processing speed to counteract motion artifacts (Pittet and Weissleder 2011; Entenberg et al. 2017, 2023; Huang et al. 2021). Further research to improve or eliminate these challenges could significantly enhance the widespread adoption and implementation of this technique.

Multiphoton microscopy is constantly being improved to provide dynamic images of the tumor and other microenvironmental cells at an unprecedented high resolution and speed. Until recently, it was not possible to image a large field of view within any given tumor, which limits the true histopathological context of the imaged cells and results in undersampling of the tissues. However, with the recent development and integration of the “large-volume, high-resolution” (LVHR) technique to in vivo imaging, it is now possible to generate a remarkably broader view of tissues than was previously possible with in vivo imaging alone (Entenberg et al. 2017). This technique produces a low-resolution mosaic image with a wide field of view by stitching multiple images with high resolution and high magnification. With the use of this technique, it is now possible to visualize and study complex biological processes such as early stages of tumor progression, metastatic spread, and their fate in a greater depth.

In vivo imaging has provided an unprecedented view of the spatial and temporal dynamics of cellular behavior within the native microenvironment in real-time. However, imaging alone cannot provide key molecular information that can be used to gain mechanistic insight into the nature of underlying cellular behaviors. Due to numerous advances, it has become increasingly possible to combine multiple-omics technologies in a limited number of cells to collect a wide range of unbiased molecular information. Successful integration of in vivo imaging with currently available high-throughput spatial transcriptomics technology such as Image-seq (Haase et al. 2022) and NICHE-seq (Medaglia et al. 2017) suggests that such -omics technology could be used to assess molecular information of the cells under investigation without destroying the underlying spatial information. Future studies based on such combinations will provide a deeper understanding of the mechanisms of intercellular communication, further advancing our understanding of cancer cell dynamics within the living environment.

COMPETING INTEREST STATEMENT

T.R. is a Co-founder and a member of the Scientific Advisory Board at Tiger Hill Therapeutics, on the Scientific Advisory Board of Styx Biotechnology, and on the Clinical Advisory Board of Signalexis and serves as a consultant for Capvision.